Conventionally, so-called tandem-type image forming apparatuses are known. This type of image forming apparatus includes a plurality of photosensitive members for each color (yellow, magenta, cyan, and black, for example) that is arranged in a direction in which a paper conveying belt moves. Images in each color borne on a corresponding photosensitive member are sequentially transferred onto a paper on the belt.
In such a tandem-type image forming apparatuses, if image forming positions on paper for each color are deviated (shifted) from the correct positions, color images with color registration errors are formed undesirably. Hence, one of these image forming apparatuses has a function to correct image forming positions of each color (Japanese Patent Application Publication No. 2004-61876). When performing this correcting function, first the image forming apparatus forms a registration pattern (position alignment pattern or calibration pattern) including a plurality of marks on the belt. The registration pattern includes reference color marks and correction color marks, and includes a plurality of pairs of marks having different degrees of overlapping between both marks in a main scanning direction or in a sub-scanning direction. Density of each pair of marks differs depending on the degree of overlapping of the pair of marks. It can be considered that the both marks constituting the pair of marks having density in an extremal value are aligned with each other (the degree of overlapping is maximum). Hence, density of each pair of marks is sequentially measured, and the pair of marks showing an extremal value of those densities is identified. Then, the positional deviation amount that is originally assumed by the image forming apparatus for the pair of marks is determined as the deviation amount of image forming positions of the correction color relative to image forming positions of the reference color in the main scanning direction or in the sub-scanning direction. The timing of forming images is corrected by offsetting the deviation amount.